Antony PRICE*, Patrick TURPIN, Michel ERBETTA, Total E&P with Don WATTS and Graeme CAIRNS, WesternGeco ElectroMagnetics. Summary Discussed here are the results of a 1D and 3D forward with 1D and 2D inverse Controlled Source Electromagnetics (CSEM) modeling exercise over a know target offshore West Africa with the intent of demonstrating the technology, develop methodology for analysis and better understanding its limitations. The 3D dataset presented here is considered rich consisting of 11 transmitter (Tx) tow lines and 43 electric and magnetic seafloor receivers (Rx) arranged in a 3D grid pattern (see Figure 1) and acquired with a time-shared dual base frequency square waveform of 0.25 and 0.0625 Hz. Aspects of detectability, comparison to modeled results (1D, 2D and 3D), and resolution of known structure both horizontal and vertical, comparison with logged resistivities in a constrained and unconstrained sense are analyzed. For example detection of the target structure seems apparent in the unconstrained 2D inversions of the CSEM data. However, this inversion has reduced ability to image separate, vertically stacked pay and laterally non-extensive bodies in the unconstrained inversions and forward modeled data. Contrast will also be made between the 1D and 2D inversions to assess the limits in imaging the hydrocarbon filled structures, along with 3D forward calculated responses to be compared with the acquired data. Combined and joint inversion of the CSEM and Marine MagnetoTelluric (MMT) information will also be considered in order to better constrain the deeper resistivity section not sampled by well logs and to improve imaging of the 3D complexity of our known target. Introduction As many descriptions of the CSEM acquisition methodology have been presented in the past, the authors will spare the audience these details. For full description of the acquisition methodology, see Elligsrud et al. 2002 or Eidesmo et al. 2002. The survey acquisition design presented many challenges as seafloor Rx and Tx tow-lines were required to respect exclusion zones for all current and future drilling and seabed infrastructure associated with the development of the field. Drop accuracy and positioning of the seafloor Rx s was required to be proven in a non-sensitive green zone before deployment in the sensitive red zone (see Figure 1) in proximity to development activities, with ROV recovery of Rx anchors planned in the case of conflict with seafloor equipment. Test were performed to demonstrate that CSEM acquisition would not interfere with drilling Figure 1: Line and reciver location with field outline in blue. Larger polygon is shalow and smaller deeper reservoir. Inset shows above plan with red zone and green zone deployment areas. operations, in particular the electromagnetic Tx energy interference with the magnetic guidance used for drilling operations and acoustic positioning of the towed Tx and Rx locations not interfering with other seafloor acoustic positioning systems. A dual frequency, time-share Tx waveform of 0.25 and 0.0625 Hz was used to broaden the bandwidth and depth range of investigation with respect to the skin depth signal penetration limitations in a conductive environment. A majority of Rx s were also equipped with vertical electrical sensors to measure the full vector compliment of the electric field. It may seem inappropriate to model such a complex dataset and target in 1D, however the ready availability of fast forward modeling and inverse code on a standard PC as well lends the 1D technique well to an initial pseudoquantitative interpretation. Indeed, Constable et al. (2006) found that 1D inversion results are appropriate if receiver Rx and Tx are over the target to be modeled. For our purposes we will argue that it is sufficient to have Rx and the common mid-point (CMP) with the Tx over our target of interest. Our initial analysis of the CSEM dataset took the form of 1D analysis in both forward and inverse modeling of the in-line electric field at multiple frequencies and harmonics. The results from the 1D analysis demonstrate a detection case where a resistive structure is 639
3D analysis initially involved forward calculation of responses generated from a detailed model constructed from all known structures seen in the seismic and resistivity logs from multiple wells. These responses were then compared to the processed amplitude and phase data to better understand the information content of the CSEM data. 3D inversion of the CSEM and an effort to joint invert the CSEM and MMT has also been considered in understanding the more subtle 3D information content and their imaging limitations. 1D analysis Figure 3: 1D forward modelled example at frequency 0.25 Hz on-target +ve domain. Calculated amplitude and phase in green, measured data in red and blue, 1D model in depth right. Figure 2: 1D inverse modelled example at frequency 0.25 Hz on-target +ve domain. Calculated amplitude and phase in green, measured data in red and blue, 1D model in depth right with fixed resisitivity below 4300m sub-sea. identified that correlates well with the known hydrocarbon accumulation. 2D inversion analysis of the in-line electrical field provided perhaps the most impressive and easily generated resistivity sections that can be co-rendered with seismic sections to clearly indicate which seismic structures produce resistivity anomalies and thus likely hydrocarbons in high saturation. Initial analysis of the dataset took the form of forward modeling to better understand not only the structure seen by the data, but also its sensitivity in terms of depth, resistivity thickness product, frequency, amplitude, phase and offset (see Figure 3). For each receiver on or near target, two models were constructed fitting the positive (blue) and negative (red) offset domain (in-tow and outtow) to account for lateral variations. As higher frequencies probe shallower depths due to skin depth limitations, the 1D resistivity forward models were constructed in a top-down, shallow to deep manner by selectively fitting the highest (1.75 Hz) to the lowest (0.25 Hz) harmonics and base frequencies. The 1D forward modeling included the rough depths of relevant seismic horizons at the CMP location, with the fit between forward calculated and observed data in these 1D models assessed as good if the difference remained below ±20% between 2 and 9 km Rx/Tx offsets. The results of the 1D modeling indicated generally higher resistivities than expected (with reference to the resistivity well logs) in the shallower section (first three layers) and lower resistivities than expected in the reservoir section, indicating some inversion ambiguity between the shallow and deep section. Indeed the shallow section is not usually logged by the well data so some uncertainty remains as to the shallowest layers resistivity. Note that little effort was made to constrain any of the resistivities with values seen in the logs at this stage as this was the unknown parameter used in the fitting of the observed data with depth of horizons roughly constrained. Interesting to note however that for modeling of receivers at the edge of the target, the shallow section (first three layers) remained roughly constant and there was a significant increase in resistivity at the reservoir level for the on-target model, as would be expected. Once a full appreciation for the sensitivity of these data at various depth, offsets and frequencies was attained by forward modeling, attention was turned to 1D inversion. Constraint was again the rough depth of the relevant seismic horizons, this time both at the Rx location and at the CMP between the Rx and Tx. This allowed comparison 640
of the effect of constraint and location on the inversion and an assessment of the best approach. Inversions were started at a simple half-space of 2.0 ohm.m for all layers. The results of the 1D inversions demonstrated an unexpected sensitivity to the airwave (as confirmed by further forward modeling) at large offsets resulting in a clear overestimation of resistors at depth, requiring a fixed resistivity to stabilize the inversion below about 4,300m meters sub-sea. The following inversions displayed a closer match with the expected resistivity profile in the subsurface, with shallow resistors being overestimated and the target resistors being underestimated. Overall the 1D inversions over the target displayed anomalous resistivity as we would expect (Figure 2). Note that the smoothness requirements of the inversion produced an integrated, smooth response over the stacked reservoir layers, demonstrating detection of the group but no definition of the individual pay zones. At this stage there was also concern over the proximity of the airwave influenced deeper structure and the resistive target and the ability to distinguish between the two. Figure 4: 2.5D unconstrained resistivity inversion (color) of an East-West CSEM line overlain on 3D seismic line (background) displaying resistivity anomaly at crest of structure and roughly correct depth and extent. 2D analysis The application of a 2-2.5D inversion (Abubakar et al 2006 constrained Gauss-Newton type) was made to illuminate additional detail in the resistivity structure, and initial results of an unconstrained inversion overlain on an equivalent seismic line extracted from a 3D seismic cube appear in Figure 4. The resistivity anomaly now has more lateral character that is associated with groups of channelized reservoirs perpendicular to the section, however once again vertical stacked pay has been grouped and smoothed in a vertical sense and less laterally extensive structures were not imaged (see Figure 5). Estimates of resistivity compared with down-sampled and smoothed well log and seismic data did however compare well as did lateral and vertical location of the resistivity anomaly. Tests of non-uniqueness were performed to establish the stability of the 2.5D inversion result which demonstrated reliability in the response at the depth range of interest. The non-uniqueness was tested by running multiple inversions from differing half-space resistivities and calculating the spread of solutions. If we also assume the resolution of the inverted results are independent of these perturbed halfspace models, then this analysis can also indicate the ultimate resolution of the inversions. Several attempts were made to constrain these inversions vertically using the available seismic horizon data as tear surfaces allowing discontinuities in the smoothness function of the inversion with no lateral restriction, with results that correlated only loosely to the known structures. It was felt at this stage that moving to a 3D modeling and perhaps inversion would better capture the complex 3D structure of our target. Figure 5: Comparison overlay of a resistivity section from the 3D model constructed from seismic and well data (wells indicated) and the 2D unconstrained inversion result from a north-south line. 3D model data display two stacked pay, 2D CSEM inversion only one rounded anomaly. 3D analysis To better understand the CSEM responses given the wealth of seismic and well data that exists over our test target, a 3D forward modeling exercise was undertaken to test various scenarios and their expression (or not) in the measured data. The various scenarios took the form of calculating a forward model response for each of the stacked reservoirs in turn from top to bottom and comparing these responses to the CSEM data to assess sensitivity to these structures (Figure 6). Note that the fit here is poor indicating that our 3D forward model is in error. It was also observed that the shallower and largest structures were more easily seen in the measured data with 641
some hint as to a screening effect of the deeper structures by the shallow, in some places more extensive resistors. Acknowledgments WesternGecoEM are acknowledged for their role in the acquisition, processing and interpretation of this rich 3D CSEM dataset, along with Total E&P for supporting this work and providing permission for the material presented here, and particularly the affiliate for their extensive technical and logistical support. Figure 6: Comparison of synthetic data from the 3D model and the measured (obs) CSEM data at the two base frequencies plotted as in-line amplitude vs Tx/Rx offset in meters. Conclusions After detailed analysis of a rich 3D CSEM dataset in 1D, 2D and 3D the methodology outlined here demonstrates utility for analyzing these data showing detection, delineation to the limits of resolution of the technique and limited structural characterization of the complex bodies that comprise this known target. Improvements in understanding of the behavior of the CSEM technique and the various modeling approaches have been made with the simpler and easier approach of 1D inversion on occasion producing useful results for the usually under-determined inversion problem. The lateral resolution however is difficult to asses given that the problem is not strictly 1D with a moving CMP depth and position of investigation. 2D inversion addresses most of these issues providing perhaps the most appropriate results for rapid analysis such as by overlay on seismic, with the normal limitation in resolution seen in these examples. Understanding developed during the 1D and 2D modeling and inversion exercise greatly aided the investigation of the 3D forward modeling. 3D inversion itself, weather joint CSEM+MMT or constrained by seismic horizons does not lend itself well to the usual problems and question of exploration due to complexity in construction and run-time on large computing clusters, however is thought to be able to provide more detailed information on the target for development and possible future monitoring purposes. 642
EDITED REFERENCES Note: This reference list is a copy-edited version of the reference list submitted by the author. Reference lists for the 2008 SEG Technical Program Expanded Abstracts have been copy edited so that references provided with the online metadata for each paper will achieve a high degree of linking to cited sources that appear on the Web. REFERENCES Abubakar, A., T. Habashy, V. Druskin, D. Alumbaugh, A. Zerelli, and L. Knizhnerman, 2006, Two-and-half-dimensional forward and inverse modeling for marine CSEM problems: 76th Annual International Meeting, SEG, Expanded Abstracts, 750 754. Constable, S., and C. J. Weiss, 2006, Mapping thin resistor and hydrocarbons with marine EM methods: insights from 1D modeling: Geophysics, 71, G43 G51. Eidesmo, T., S. Ellingsrud, L. M. MacGregor, S. Constable, M. C. Sinha, S. Johansen, F. N. Kong, and H. Westerdahl, 2002, Sea bed logging (SBL): A new method for remote and direct identification of hydrocarbon filled layers in deepwater areas: First Break, 144 152. Ellingsrude, S., T. Eidesmo, S. Johansen, M. C. Sinha, L. M. MacGregor, and S. Constable, 2002, Remote sensing of hydrocarbon layers by seabed logging (SBL): Results from a cruise offshore Angola: The Leading Edge, 972 982. 643